News Intel spins off Network and Edge group as standalone business after posting $2.9b loss — seeking investors for NEX division amid massive internal r...

[bit_user]

If they don't get customers, which is extremely unlikely since they got customers even for 18A....

Then why would he be threatening it? He also said a "major" customer, meaning probably someone like Apple or Nvidia.

Furthermore, saying 14A is on the threshold of being cancelled is a horrible message, if you're trying to attract customers. That might even scare away any who were considering it.

18A wouldn't disappear though, they would still use that for many years, same as all the older functional FABs.
They would also not get rid of plots of lands, what would be the reason?!

They can only use 18A as long as the chips made on it are competitive, and when sold at a price that doesn't lose money.

They would get rid of the land because they need money. @thestryker , they just Intel just recently sell off its Oregon campus and switch to leasing it (or something like that)?

 

[TerryLaze]

Then why would he be threatening it? He also said a "major" customer, meaning probably someone like Apple or Nvidia.

Furthermore, saying 14A is on the threshold of being cancelled is a horrible message, if you're trying to attract customers. That might even scare away any who were considering it.

Nobody cares why he said it!
If 14A makes money it will make money(and intel will build the fabs) if it doesn't then it won't (and they won't build the fabs) ,how difficult do you have to make this extremely simple thing????

They can only use 18A as long as the chips made on it are competitive, and when sold at a price that doesn't lose money.

Have you been around for 14++++++++ ?
Trick question, we all know you have.
They can make enough money from 18A to found their next node, they just won't be able to keep up their tick-tock model, instead releasing new nodes slower.

If they give up being an external foundry then they don't need to have a new node every year.

 

[thestryker]

It's funny how Nvidia and AMD recently moved in the opposite direction. Nvidia bought Mellanox about 6 years ago, and AMD bought Pensando like 3 years later. That's because they both understand how essential networking is to scale AI training.

Intel failing to buy Mellanox seems to have been a really big problem. The relative stagnation of the networking side really made me wonder if they'd been under investing. When you look at the products that have made it to market pretty much every other company has something better in every segment. Intel doesn't even have a 5Gb controller so is renaming a Realtek one for their Killer integrated.

Their current 10GBase-T chips run too hot and really need active cooling for a proper consumer product.

Their current ones (E610) are actually the best on the market it just took them until this year to get there!

 

[thestryker]

If they don't get customers, which is extremely unlikely since they got customers even for 18A....

They don't have any big/ongoing orders for 18A and that has driven the stance on 14A. Tan is basically saying that if 14A is like 18A then IFS 2.0 is dead. That would inevitably end with Intel selling off the fabs as soon as financially viable. It does seem like Intel is already setting up for this future with the way they've been replacing equipment early and firing fab engineers.

 

[thestryker]

Item 3 is a massive uphill battle, given how far behind Intel is on AI. AMD is much further along and still way behind Nvidia.

I've thought about this a bunch the last couple of years and came to the conclusion that if Intel really wanted a piece of that AI pie IFS was probably the way to do it. That possibility went out the window with Gelsinger.

They can only use 18A as long as the chips made on it are competitive, and when sold at a price that doesn't lose money.

Intel 3 and 18A (and their derivatives of course) will both be forever nodes because the EUV fabs need to be filled as much as possible until they can be sold.

They would get rid of the land because they need money. @thestryker , they just Intel just recently sell off its Oregon campus and switch to leasing it (or something like that)?

They haven't done that in Oregon and I don't think they can afford to given that I think they have fab equipment at every one large enough the sell/lease scheme would make sense. They did it down in California though.

 

[logainofhades]

It looks like it was CA and not OR.

https://www.tomshardware.com/pc-com...-capital-but-maintaining-operations-and-staff

 

[Giroro]

Rip and tear until Intel's done.

 

[bit_user]

Nobody cares why he said it!

A CEO doesn't just threaten to cancel such a huge part of their business on a whim. He must've said it as fair warning to investors that it could realistically happen. That's the only reason he would say such a thing.

If 14A makes money it will make money(and intel will build the fabs) if it doesn't then it won't (and they won't build the fabs) ,how difficult do you have to make this extremely simple thing????

It's not about building fabs. 14A would happen in fabs they've already built. The statement was that if 14A doesn't get a major customer, it and all future nodes will be cancelled. In other words, Intel will get out of the fabrication business. That's how dire it is.

 

[bit_user]

Their current ones (E610) are actually the best on the market it just took them until this year to get there!

This dual-port model has a list price of ~$250 ??

https://www.intel.com/content/www/u...t-network-adapter-e610xt2/specifications.html
​

How can you even charge that for 10G, in this day and age?

The difference in list price between my motherboard and the version without 2x 10G ports integrated was only about $100, which roughly tracks with what you can find 10G PCIe NICs for, on the open market.

 

[TerryLaze]

It's not about building fabs. 14A would happen in fabs they've already built. The statement was that if 14A doesn't get a major customer, it and all future nodes will be cancelled. In other words, Intel will get out of the fabrication business. That's how dire it is.

You mean your statement was....
The actual statement leaves everything completely open and doesn't close any door on anything.
But if you would boil it down to a simple sentence his statement says: If it will make money we will do it, if it doesn't, we'll have to see.
https://www.intc.com/filings-report...0000050863-25-000109/0000050863-25-000109.pdf

We are focused on the continued development of Intel 14A, the next generation node beyond Intel 18A and Intel 18A-P, and on securing a significant external customer for such node.

However, if we are unable to secure a significant external customer and meet important customer milestones for Intel 14A, we face the prospect that it will not be economical to develop and manufacture Intel 14A and successor leading-edge nodes on a go-forward basis.

In such event, we may pause or discontinue our pursuit of Intel 14A and successor nodes and various of our manufacturing expansion projects.

MAY
....
PAUSE
....
or discontinue

While we continue to evaluate Intel 14A for use in future Intel products and our plan includes an initial product designed to utilize Intel 14A, at present we are maintaining the option to design future Intel products requiring nodes with performance beyond Intel 18A and Intel 18A-P to be produced internally or by an external foundry.

If we were to discontinue development of Intel 14A and successor nodes, we expect that a majority of our products would continue to be manufactured in our own facilities utilizing our nodes up to Intel 18A-P through at least 2030.

If 14A gets cancelled they will still milk 18A till the end of days....of 2030

By focusing on our customers and delivering the best semiconductor products to the market, manufactured on the most appropriate internal or external node from a performance and cost perspective,

and only deploying capital on new nodes and manufacturing facilities where we believe they will yield an attractive return, we believe we can improve the competitiveness of our products business, and the overall financial results for the company.

Whenever new node gets affordable new node will be made.

 

[thestryker]

This dual-port model has a list price of ~$250 ??

I didn't say it made sense to buy! 🤣

They list the node on the X710 as 28nm and the E610 as 7nm so I feel like they could have done literally nothing architecture wise and it still would have been significantly more efficient. This would also imply that they're not using in house nodes so maybe that's another reason why the hammer got dropped on NEX.

 

[bit_user]

The actual statement leaves everything completely open and doesn't close any door on anything.
But if you would boil it down to a simple sentence his statement says: If it will make money we will do it, if it doesn't, we'll have to see.
https://www.intc.com/filings-report...0000050863-25-000109/0000050863-25-000109.pdf

Oh, it definitely slams the door shut. Once you take your foot off the gas, you're following the Global Foundries path, with no real way to catch up.

Whenever new node gets affordable new node will be made.

This is nonsense. You cannot stop and start new node development like turning off and on a faucet. It takes a huge organization and loads of expertise and institutional knowledge. If they stop 14A and further node development, they won't keep those people around and all of that capability to do node development drains away almost instantly.

 

[thestryker]

Whenever new node gets affordable new node will be made.

I just want to highlight this, because it's very wrong. Unless Intel is licensing nodes from someone else this is 100% not a viable strategy (even then it is unlikely to be). They can continue to refine Intel 3 and 18A improving them and even making custom optimized versions of them for customers, but a new node is not something you can flip on and off.

 

[Giroro]

I think Tan clearly laid it out in his email to employees--really to all stakeholders, since it was published on the Intel website. His "three strategic pillars of growth" are 1) get foundry capex under control, 2) revitalize x86 (the current breadwinner), and 3) do AI (the future breadwiner).

Here's the problem with those pillars: they aren't pillars. They work against each other, not together. They're mutually exclusive.

It's impossible for Intel to cut capex on foundry and accomplish the other two goals at the same time - because foundry was always their primary competitive advantage. They can never compete with Nvidia/Apple/AMD if they are planning to show up a year late to the party to buy TSMC's leftover capacity - and they can't afford to be first in line. Even if they could, that would mean a far more expensive unit cost than when they do it in house.

x86 vs AI is also a problem. If Intel focused their priority on either one, it would pull customers away from the other. That doesn't necessarily have to be true if they were growing rapidly and had unlimited resources to find some synergy to boost both businesses. The two options don't have to be mutually exclusive, a company with leadership in one could afford to become the leader of both.
Going back to the importance of foundry capex - if tomorrow some company with mediocre designs like MediaTek or VIA jumped in and magically had a way to build silicon 10x better than what TSMC has, then any garbage design they are selling would become the leader in both categories overnight. It's technically possible for one company to be the best in both general compute and AI.
But Intel doesn't get to become the best at both, because Intel is in a death spiral. Their lack of current technical leadership (and stated unwillingness to spend what it takes to regain that leadership) means they won't win new business. Intel has an existing pool of customers, who are sticking around primarily for convenience - and will eventually run out. It's right there in the press release, they will only build to meet existing demand.
If Intel's foundries jumped ahead to near-parity with TSMC, then they might have the resources to regain lead in either general compute or AI.
But trying to become leader in both categories at the same time, while also trying to cut costs, sell-off any potential untapped technical advantages, and share a foundry with your competitors?
Absolutely impossible. Impossible isn't a strong enough word. It's a ridiculous approach to bleeding edge. We're talking about a ~$90 Billion company who's going to need to outbid Nvidia, who's CEO alone has a personal net-worth over $140 Billion. If Intel ever manages to make anything even remotely threatening to Nvidia, they they're going to get eaten.

I think CEO Lip-Bu Tan was put into place because he knows how to help the board cash-out: De-clutter the junk and put the house on the market. I can't say for sure what the real strategy is, but to me I don't think that Lip-Bu Tan is a complete idiot. I just think he was hired because he is incredibly well-connected with potential buyers of the company, and he knows how to (legally) rob investors.
Or maybe they hope that Musk, Ellison, or Huang will get bored and just buy the brand to hang on their wall as some kind of weirdo hunting trophy.

 

[-Fran-]

Mr. Lip Bu Tan going nuclear is... Funny.

EDTI: I don't have anything new or interesting to add, as most of the good discussion has already taken place. Be amused (or not) by my bad joke as a replacement.

Regards.

 

[Moores_Ghost]

He did say few days ago that he wants to change the company culture, but he didn't say how he is going to achieve that. Even if he does have a plan, I doubt it is doable. The toxic culture is such a big part of Intel's existence, I don't think it can be changed. He would need to fire everyone and hire new people, which is not a solution.

Why isn't that a solution? He's already doing that. One division at a time. Since he's been in charge, too.

 

[Moores_Ghost]

It's funny how Nvidia and AMD recently moved in the opposite direction. Nvidia bought Mellanox about 6 years ago, and AMD bought Pensando like 3 years later. That's because they both understand how essential networking is to scale AI training.

Even though Intel will still retain part ownership, what they lose by bringing in other investors is that the entity must now prioritize its own interests. Intel can no longer twist its product roadmap or pricing structure to support its efforts to sell its servers or AI accelerators in ways that might not maximize shareholder value for the new company.

For instance, if the new company wants to work with another AI chip company, like Tenstorrent, I think Intel can't block that, even with a majority stake, unless they want to open themselves up to litigation by other investors claiming that Intel isn't making decisions in the new company's best interests.

What's funnier is that back before covid, Jim Keller explained this in an interview. Gotta love Jim. His DNA is in everything.